UNIVERSITI PUTRA MALAYSIA IMAGE COMPRESSION BASED ON REGION OF INTEREST FOR COMPUTERIZED TOMOGRAPHY IMAGES TARIK FARAJ ALI IDBEAA FK

Transcription

1 UNIVERSITI PUTRA MALAYSIA IMAGE COMPRESSION BASED ON REGION OF INTEREST FOR COMPUTERIZED TOMOGRAPHY IMAGES TARIK FARAJ ALI IDBEAA FK

2 IMAGE COMPRESSION BASED ON REGION OF INTEREST FOR COMPUTERIZED TOMOGRAPHY IMAGES By TARIK FARAJ ALI IDBEAA Thesis Submitted to the School of Graduate Studies, University Putra Malaysia in Partial Fulfillment of the Requirements for the Degree of Master of Science March 2003

3 To my parents, wife, daughter, Brothers and sister n

4 Abstract of thesis presented to the Senate of Universiti Putra Malaysia in partial fulfilment of the requirements for the degree of Master of Science. IMAGE COMPRESSION BASED ON REGION OF INTEREST FOR COMPUTED TOMOGRAPHY IMAGES By TARIK FARAJ ALI IDBEAA March 2003 Chairman: Abdul Rahman Ramli, Ph.D. Faculty : Engineering The use of computers for handling image data in the healthcare is growing. The amount of data produced by modem image generating techniques, such as Computed Tomography (CT) and Magnetic Resonance (MR), is vast. The amount of data might be a problem from a storage point of view or when the data is sent over a network. To overcome these problems data compression techniques adapted to these applications are needed. Many classes of images contain some spatial regions which are more important than other regions. Compression methods which are capable of achieving higher reconstruction quality of important parts of the image have been implemented. For medical images, only a small portion of the image might be diagnostically useful, but the cost of wrong interpretation is high. Algorithms which deliver lossless compression within the regions of interest (ROJ), and lossy compression elsewhere 111

5 in the image, might be the key to providing efficient and accurate image coding to the medical community. In this thesis both of compression techniques (lossy and lossless) of medical images using the JPEG algorithm (DCT), will be discussed. IV

6 Abstrak tesis yang dikemukakan kepada Senat Universiti Putra Malaysia sebagai memenuhi keperluan untuk ijazah Master Sains. IMEJ PEMAMPATAN BERASASKAN BAHAGIAN YANG MENARIK UNTUK IMEJ TOMOGRAFI BERKOMPUTER Oleh TARIK FARAJ ALI IDBEAA Mac 2003 Pengerusi: Abdul Rahman Ramli, Ph.D. Fakulti : Engineering Penggunaan komputer untuk mengendalikan data imej di dalam penjagaan kesihatan sedang berkembang. Jumlah data yang dihasilkan oleh teknik moden timbulan imej, seperti CT dan MR, sangat besar. Jumlah data mungkin menjadi suatu masalah daripada segi penyimpanan atau bila data dihantar menerusi sesuatu rangkaian. Untuk mengatasi masalah ini, teknik pemampatan data digunakan. Banyak kelas imej mengandungi beberapa buah kawasan lapang yang mana lebih penting berbanding dengan kawasan lain. Kaedah pemampatan yang boleh mencapai kualiti pembinaan semula yang lebih tinggi telah dilaksanakan. Bagi imej perubatan, mungkin sebahagian kecil imej sahaja berguna untuk diagnosis, tetapi harga bagi salah tafsiran sangat tinggi. Algoritma yang mana menyampaikan pemampatan tanpa hilang dalam kawasan yang menarik (RGI), dan pemampatan hilang lain tempat di dalam imej, mungkin menjadi kunci untuk membekalkan v

7 gambaran asal dan tepat kepada masyarakat perubatan. Dalam tesis ini kedua-dua teknik pemampatan (hilang dan tanpa hilang) dilaksanakan dan dibincangkan ke atas imej-imej perubatan yang menggunakan algoritma JPEG (DCT). VI

8 ACKNOWLEDGEMENTS I would like to express my appreciation to Dr. Abdul Rahman Ramli during this work. His patience and guidance have been invaluable to me. Without the time he has invested brainstorming ideas and discussing results, this project would never have gotten off the ground. However, this is not the most important thing he has done for me. Usually, I see him demonstrating a love for teaching and learning. He has supported me in every way through both the easy and the difficult times. I count myself fortunate to have known and worked with Dr. Abdul Rahman Ramli. I would also like to thank my committee members Dr. Veeraraghavan Prakash and Puan Roslizah Ali. And last, but not least, I would like to thank my family and especially my wife for helping me during my studying. Their help and support throughout the years has made it possible for me to come this far. I am sure that they had no idea what I was talking about half the time, but when things got difficult they were there to let me talk things through and rejoiced with my success. Vll

9 I certify that an Examination Committee met on 20 th March 2003 to conduct the final examination of Tarik Faraj Ali Idbeaa on his Master of Science thesis entitled "Image Compression Based on Region of Interest for Computerized Tomography Images" in accordance with Universiti Pertanian Malaysia (Higher Degree) Act 1980 and Universiti Pertanian Malaysia (Higher Degree) Regulations The Committee recommends that the candidate be awarded the relevant degree. Members of the Examination Committee are as follows: Shattri Mansor, Ph.D. Associate Professor, Faculty of Engineering, Universiti Putra Malaysia. (Chairman) Abdul Rahman Ramli, Ph.D. Faculty of Engineering, Universiti Putra Malaysia. (Member) Veeraraghavan Prakash, Ph.D Faculty of Engineering, Universiti Putra Malaysia. (Member) Roslizah Ali, M.S. Faculty of Engineering, Universiti Putra Malaysia. (Member) GULAM US L RAHMA T ALI, Ph.D. Professor / Deputy Dean, School of Graduate Studies, Universiti Putra Malaysia. Date: 2 J APH / 03 Vlll

10 The thesis submitted to the Senate of Universiti Putra Malaysia has been accepted as fulfillment of the requirement for the degree of Master of Science. The members of the Supervisory Committee are as follows: Shattri Mansor, Ph.D. Associate Professor, Faculty of Engineering, Universiti Putra Malaysia (Chairman) Abdul Rahman Ramli, Ph.D. Faculty of Engineering, Universiti Putra Malaysia (Member) Veeraraghavan Prakash, Ph.D Faculty of Engineering, Universiti Putra Malaysia (Member) Roslizah Ali, MS. Faculty of Engineering, Universiti Putra Malaysia (Member) AINI IDERIS, Ph.D. Professor / Dean School of Graduate Studies, Universiti Putra Malaysia. Date: 12 JUN 2003 lx

11 DECLARA TION I hereby declare that the thesis is based on my original work except for equations and citations, which have been duly acknowledged. I also declare that it has not been previously or concurrently submitted for any other degree at UPM or other institutions. J ALI IDBEAA) Date: Z y. () ' '3 x

12 TABLE OF CONTENTS Page DEDICATION ABSTRACT ABSTRAK ACKNOWLEDGEMENTS APPROVAL DECLARATION LIST OF TABLES LIST OF FIGURES LIST OF ABBREVIATIONS ii 111 v VII VIn x XIV XVI xix CHAPTER I. INTRODUCTION 1 Introduction to Medical Image Compression 1 ws 4 Thesis Organization 5 II. LITERATURE REVIEW 6 Image Compression in Healthcare 6 Radiology 6 Cardiology 7 Medical Image Distortion And Quality 8 Image Quality Measurements 8 Image Artifacts 9 The Human Visual System 11 Medical Image Standards 11 DICOM (Digital Imaging and Communications in Medicine) 12 PACS (Picture Archiving and Communication System) 12 JPEG (Joint Photographic Expert Group) 13 EPIC Compression Algorithm 15 APMAR Model for Medical Image Compression 16 Storing Medical Image 17 Regions Fundamentals and Segmentation 19 Regions with Different Encoding Methods 20 Regions for Motion Compensation 21 Regions with Different Quality Levels 22 Region of Interest Based Compression of Medical Images 22 Segmentation Using Morphology 24 Region of Interest (ROI) Based Compression System 25 Medical Image Data Formats 28 Image Data Compression Schemes Lossless Image Compression 30 Lossy Compression Xl

13 Coding Techniques 36 Huffman Coding 36 Runlength Coding 36 Arithmetic Coding 37 Entropy Coding ( Lempel - Ziv ) 37 Delta Encoding 37 Lossless Differential Pulse Code Modulation 38 Block-Coding / Pixel Block Compression 38 Area Coding 38 Lossy Differential Pulse Code Modulation Truncation Coding Lossy Pixel Block-Coding 40 Vector Quantization 40 Fractal Image Compression 41 DCT - Based Transform Coding 41 Wavelet - Based Image Compression 42 Compression Of Digital Medical Images 42 Compression Of Digital Medical Images Due to the Classification 43 Of the Image Bit Plane Adaptive Compression Of Digital Medical Images With out loss 45 Of Information JPEG Standards 49 Structure of the Standard 50 Summary III. IV. METHODOLOGY Introduction Compression Efficiency Mathematical Evaluation Choosing Software Tool System Methodology Reading Gray Level Medical Images Select the Region of Interest Lossy Compression Lossless Compression Combine Results Store the Compressed Image RESULTS AND DISCUSSION Introduction Medical Image Compressed in A case of ROI size (l0ixi01) Quantization Multiplier Factor = 2 Quantization Multiplier Factor = 5 Quantization Multiplier Factor = 10 Quantization Multiplier Factor = 20 Quantization Multiplier Factor = 40 Medical Image Compressed in A case of ROI size (51x51) Quantization Multiplier Factor = 2 Quantization Multiplier Factor = 5 Quantization Multiplier Factor = xu

14 Quantization Multiplier Factor = 20 Quantization MUltiplier Factor = V. CONCLUSION AND RECOMMENDATIONS Conclusion Recommendations REFERENCES BIODATA OF THE AUTHOR xiii

15 LIST OF TABLES Table Page 4.1 Lossless regions of interest results with size (101 x 101) Lossy compression results without ROI (101 x 101) and MF = Final output compressed image with (101 x 101) lossless region 82 of interest and MF = Lossy compression results without ROI (101 x 101) and MF = Final output compressed image with (101 x 101) lossless region 84 of interest and MF = Lossy compression results without ROI (101 x 101) and MF=lO Final output compressed image with (101 x 10 1) lossless region 86 of interest and MF = Lossy compression results without ROI (101 x 101) and MF= Final output compressed image with (101 x 101) lossless region 88 of interest and MF = Lossy compression results without ROI (101 x 101) and MF= Final output compressed image with (101 x 101) lossless region 90 of interest and MF = Lossless regions of interest results with size (51 x 51) Lossy compression results without ROI (51 x 51) and MF = Final output compressed image with (51 x 51) lossless region of interest and MF = Lossy compression results without ROI (51 x 51) and MF = Final output compressed image with (5 1 x 51) lossless region of 97 interest and MF = Lossy compression results without ROI (51 x 51) and MF = Final output compressed image with (51 x 51) lossless region of 99 interest and MF = 10 xiv

16 4.19 Lossy compression results without ROI (51 x 51) and MF = Final output compressed image with (5 1 x 51) lossless region of 101 interest and MF = Lossy compression results without ROI (5 1 x 51) and MF == Final output compressed image with (51 x 51) lossless region of 103 interest and MF == 40 xv

17 LIST OF FIGURES Figure Page 1.1 An MR brain scan with cancerous tumors circled JPEG encoder and decoder block diagrams Encoder of the APMAR A schematic of ROI Reconstruct CT slices with the original images Segmenting the colon from the CT data set Block diagram of the ROI based compressor ROI based compression results with block size 16x ROI based compression results with block size 8x Image compression schemes A general block diagram of DPCM A block diagram of the predictive 2-D MAR method Lossy image compression scheme Fractals in image Test images on which the algorithm "classification of the 44 bit plane is applied 2.15 Shows a flow diagram of the adaptive compression of digital medical images Three test images divided in three different classes 48 (K1,K2 and K3) after applying the classification algorithm 2.17 Block diagram of the JPEG General Methodology form 54 XVI

18 3.2 Flowchart of the encoding process Flowchart of the decoding process Flowchart of Lossy encoder Noise reduction Flowchart of Lossy Decoding Flowchart of lossless Original medical images The compressed ROI for the three images 79 (Brain, Abdomen, and Chest) with size of (101 x 101) 4.3 The three compressed images with (101 x 101) 80 Lossless compressed region of interest and MF=2 4.4 The three compressed images with the lossless 83 Compressed region of interest and MF= The three compressed images with (101 x 10 1) 85 lossless compressed region of interest and MF= The three compressed images with (101 x 101) 87 lossless compressed region of interest and MF= The three compressed images with (101 x 101) 89 lossless compressed region of interest and MF= MSE against Multiplier factor values for CT brain, chest 91 and abdomen images with out ROI (101x101) 4.9 PSNR against Multiplier factor values for CT brain, chest and 91 abdomen images with out ROI (10Ix101) The compressed ROI for the three images (Brain, Abdomen, 92 and Chest) The three compressed images with (51 x 51) 10ssless compressed 93 region of interest and MF = The three compressed images with (51 x 51) lossless compressed 96 region of interest and MF = The three compressed images with (51 x 51) lossless compressed 98 region of interest and MF = 10. XVll

19 4.14 The three compressed images with (51 x 51) lossless compressed 100 region of interest and MF= The three compressed images with (51 x 51) lossless compressed 102 region of interest and MF= MSE against Multiplier factor values for CT brain, chest and 104 abdomen images with out ROI (51x51) PSNR against Multiplier factor values for CT brain, chest and 104 abdomen images Compression ratio against ROI size for CT brain, abdomen and 105 chest images Image file size (KB) against MF for CT brain, abdomen and 106 chest Image with ROI size of (lolxlol) 4.20 Image file size (KB) against MF for CT brain, abdomen and 107 chest Image with ROI size of (5 1x51) 4.21 Image intensity variance against MF for CT brain, abdomen and 108 chest Images with ROI size (l01xi01) 4.22 Image intensity variance against MF for CT brain, abdomen and 109 chest Images with ROI size (51x51) 4.23 Comparison of PSNR value between ROI, rest of the image and 110 final compressed image with MF =2 and ROI (101x101) 4.24 Comparison of PSNR value between ROI, rest of the image and 110 final compressed image with MF =40 and ROI (l01x101) 4.25 Comparison of PSNR value between ROI, rest of the image and 111 final compressed image with MF =2 and ROI (51x51) 4.26 Comparison of PSNR value between ROI, rest of the image and 112 final compressed image with MF =40 and ROI (51x51) XVlll

20 LIST OF ABBREVIATIONS MR CT ROI CR MAE ACC ACR NEMA SNRs PSNR MSE DPCM MF EPIC ROC DEI BPE MAR RLE DCT DWT VQ SVQ Magnetic Resonance Computerized Tomography Region Of Interest Compression Ratio Maximum Absolute Error American College of Cardiology American College of Radiology National Electrical Manufacturers Association Signal-to-Noise Ratios Peak Signal-to-Noise Ratio Mean Square Error Differential Pulse Code Modulation Multiplier Factor Efficient Protocol Independent Compression Receiver Operating Characteristic Displacement Estimated Interframe Bit-Plane Encoding Multiplicative Autoregression Run-Length Encoding Discrete Cosine Transforms Discrete wavelet transform Vector quantization Scalar-Vector Quantizes XIX

21 VL C-ECSQ HVS CU RMSE Variable Length Code - Entropy Coded Scalar Quantization Human Visual System Coding Unit Root mean square error ACR-NEMA American College of Radiology - National Electric Manufacturers Association DICOM ISO PACS RIS NLIVQ HIS SS JPEG ISDN APMAR Digital Imaging and Communications in Medicine International Organization of Standardization Picture Archiving and Communication Systems Radiology Information System Non-linear interpolative vector quantization Hospital Information System Scheduling System The Joint Photographic Expert Group Integrated Services Digital Network Adaptive predictive multiplicative autoregressive ACC-NEMA American College of Cardiology - National Electric Manufacturers Association xx

22 CHAPTER I INTRODUCTION 1.1 Introduction to Medical Image Compression Today a lot of hospitals handle their medical image data with computers. The use of computers and a network makes it possible to distribute the image data among the staff efficiently. As the health care is computerized new techniques and applications are developed, among them the Magnetic Resonance (MR) and Computerized Tomography (CT) techniques. MR and CT produce sequences of images (image stacks) each a cross-section of an object. The amount of data produced by these techniques is vast and this might be a problem when sending the data over a network. To overcome this, image data have to be compressed. For two-dimensional data there exist many compression techniques such as JPEG, GIF and the new wavelet based JPEG2000 standard. All of the schemes are used for two-dimensional data (images) and while they are excellent for images, they might not be that well suited for compression of three-dimensional data such as image stacks. The easy, rapid, and reliable digital transmission and storage of medical and biomedical images would be a tremendous boon to the practice of medicine. Patients in rural areas could have convenient access to second opinions. Patients readmitted to hospitals could have earlier imaging studies instantly available. Rather than waiting for others to finish with hardcopy films, medical and surgical teams collaborating on patient care could have simultaneous access to imaging studies on monitors throughout the hospital. This long-term digital archiving or rapid transmission is prohibitive without the use of image compression to reduce the file sizes. For 1

23 example, a single analog mammogram might be digitized at 4096 x 4096 pixels x 16 bpp. This file would be over 33 megabytes (MB). In lossless compression, the original image is exactly recoverable from the compressed format; with lossy coding, it is not, but vastly greater compression is achieved. However, lossy schemes are viewed with suspicion by many members of the medical and scientific community; image alteration might entail loss of diagnostic or scientific utility. Many physicians feel they cannot trust lossy compression which mostly delivers exquisite quality and yet which can, without warning, introduce medically unacceptable artifacts into the image. After segmenting an image into regions (either automatically or manually) it is possible for a compression algorithm to deliver different levels of reconstruction quality in different spatial regions of the image. One could accurately (losslessly) preserve the features needed for medical diagnosis or for scientific measurement, while achieving high compression overall by allowing degradation in the unimportant regions. In radiology, the discussion of image compression often divides into three separate uses: compression before primary diagnosis (for rapid transmission), compression after primary diagnosis (for long-term archiving), and compression for database browsing (where progressivity would be useful)(jacob and Pamela, 1996). Compression occurring before primary diagnosis is the most controversial use of lossy compression. However, it might prove useful in cases where the interpreting radiologist is at a remote site and lossless compression cannot be used. For example, the patient's situation might require such rapid action that the time for lossless 2

24 transmission of original images cannot be countenanced, or the bandwidth for realtime lossless video transmission might not be available. Compression after primary diagnosis might be useful for long-term digital archiving. Here it is easy to imagine how region-based coding might play a role, since the primary interpretation of a film can perhaps be used for providing the region segmentation. A third use of compression is for providing progressive transmission capabilities when receiving images over a network. With progressive coding, image quality incrementally improves as more bits arrive. Early versions of an image can be good enough to show that the image is not of interest; transmission can then be 'nipped in the bud.' Progressive codes can be designed to be eventually lossless, so that if the user waits long enough (e.g., 30 seconds) the image will be exactly equal to the original, but over the short term (e.g., 0.5 seconds) the image would already be useful version more rapidly. As an example, Figure 1.1 shows an MR brain scan with cancerous tumors circled. Following its use in primary diagnosis, this image could be compressed so as to perfectly preserve this region. Figure 1.1: (a) MR brain scan with tumors (circled),(b) MR brain scan compressed by a factor of 100: 1, (c )compressed scan with circled region shown at original accuracy. 3

25 Allowing graceful degradation in the rest of the image could yield high compression. The original image has a grayscale resolution of 8 bits per pixel (bpp); Figure 1.1 b shows it compressed to 0.08 bpp. Figure 1.1c shows an example, in which the compressed image has the circular tumor region at original quality, and it represents what one might wish to obtain from a regionally lossless scheme. This image is suitable for comparison and still provides dramatically higher compression than can be achieved by schemes which are lossless everywhere. 1.2 Objectives This thesis presents a solution method to medical image compression that aims to achieve a good quality for the area that contains important information while achieving a higher compression for unrequired regions. Regions are compressed using a lossless technique; while the latter regions are compressed using a lossy technique.lossless compression reproduces the original image exactly, unlike lossy compression which trades off higher compression with slight reproduction errors. The objectives of this thesis are: 1. To implement a lossless technique for medical image compression for the ROI. 2. To investigate the effects of the multiquantization to the image. 3. To display the result of the image and evaluate the performance of selected algorithm. 4